Studies on the key methods for compressive ghost-image tracking based on background subtraction

Abstract

Abstract. Efficient object tracking represents a technology important for many vision applications. It is known that ghost imaging (GI) has a great potential if compared with a standard imaging and solves many problems in case if the common object tracking cannot be carried out. Here we show how the techniques of compressive GI and background subtraction can achieve object tracking. First, object information is captured with the GI. A characteristic measured for an object is obtained by subtracting background in the compressed domain. This characteristic uses compressive sensing to reconstruct the object image. Then the object image is projection-positioned to obtain the corresponding centroid coordinates. At last, the object trajectory is recovered with a polynomial fit, thus providing successful object tracking. Our simulation experiments suggest that the technique can track objects accurately under condition of low sampling ratios. Moreover, it decreases drastically the number of measurements needed for reconstruction and improves the tracking efficiency.